test

Cargo.lock

@@ -454,6 +454,12 @@ version = "0.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2304e00983f87ffb38b55b444b5e3b60a884b5d30c0fca7d82fe33449bbe55ea"
 
+[[package]]
+name = "hound"
+version = "3.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "62adaabb884c94955b19907d60019f4e145d091c75345379e70d1ee696f7854f"
+
 [[package]]
 name = "ident_case"
 version = "1.0.1"
@@ -1288,6 +1294,7 @@ name = "sbv2_core"
 version = "0.1.0"
 dependencies = [
  "anyhow",
+ "hound",
  "jpreprocess",
  "ndarray",
  "once_cell",

sbv2_core/Cargo.toml

@@ -5,6 +5,7 @@ edition = "2021"
 
 [dependencies]
 anyhow.workspace = true
+hound = "3.5.1"
 jpreprocess = { version = "0.10.0", features = ["naist-jdic"] }
 ndarray = "0.16.1"
 once_cell = "1.19.0"

sbv2_core/src/error.rs

@@ -16,6 +16,8 @@ pub enum Error {
     SerdeJsonError(#[from] serde_json::Error),
     #[error("IO error: {0}")]
     IoError(#[from] std::io::Error),
+    #[error("hound error: {0}")]
+    HoundError(#[from] hound::Error),
 }
 
 pub type Result<T> = std::result::Result<T, Error>;

sbv2_core/src/model.rs

@@ -1,7 +1,26 @@
 use crate::error::Result;
+use hound::{SampleFormat, WavSpec, WavWriter};
 use ndarray::{array, Array1, Array2, Axis};
 use ort::Session;
 
+fn write_wav(file_path: &str, audio: &[f32], sample_rate: u32) -> Result<()> {
+    let spec = WavSpec {
+        channels: 1, // モノラルの場合。ステレオなどの場合は2に変更
+        sample_rate,
+        bits_per_sample: 16,
+        sample_format: SampleFormat::Int,
+    };
+
+    let mut writer = WavWriter::create(file_path, spec)?;
+    for &sample in audio {
+        let int_sample = (sample * i16::MAX as f32).clamp(i16::MIN as f32, i16::MAX as f32) as i16;
+        writer.write_sample(int_sample)?;
+    }
+    writer.finalize()?;
+
+    Ok(())
+}
+
 pub fn synthesize(
     session: &Session,
     bert_ori: Array2<f32>,
@@ -25,5 +44,8 @@ pub fn synthesize(
         "bert" => bert,
         "ja_bert" => style_vector,
     }?)?;
+
+    let audio_array = outputs.get("output").unwrap().try_extract_tensor::<f32>()?;
+    write_wav("output.wav", audio_array.as_slice().unwrap(), 44100)?;
     Ok(())
 }

sbv2_core/src/utils.rs

@@ -1,4 +1,3 @@
-
 pub fn intersperse<T>(slice: &[T], sep: T) -> Vec<T>
 where
     T: Clone,